EFF urges courts not to rewrite copyright law around generative AI hype

The Electronic Frontier Foundation is pushing back on attempts to reshape copyright law around generative AI, arguing that courts are being asked to abandon long-standing principles on the basis of speculation rather than evidence.

In a post published on its Deeplinks blog on August 31, the digital rights group says rightsholders in lawsuits across the US and elsewhere are requesting exactly what the US Supreme Court warned against during the videotape era: a dramatic expansion of copyright driven by hype. The piece subsequently surfaced on the front page of Hacker News.

Two cases, one theory

EFF says it has filed amicus briefs in two of the ongoing disputes: Concord Music Group, Inc. v. Anthropic PBC and In re Mosaic LLM Litigation. Its target in both is what it calls the "market dilution" theory — the claim that because generative AI tools might unleash a flood of competing works onto creative markets, merely building those tools cannot qualify as fair use.

According to EFF, that reasoning gets copyright backwards. The system is designed to punish infringement, not competition, and to encourage others to build on existing works. Accepting market dilution, the group argues, would gut fair use and other doctrinal limits that stop rightsholders from claiming ownership over tropes, genres and styles. Publishers would gain what EFF describes as unchecked veto power over any expression that might conceivably compete with something they own — a result that would leave the public with less art and fewer ideas, regardless of whether AI is involved.

A recurring panic

EFF situates the AI backlash in a familiar historical pattern. In the 1980s, film industry figures described home video recording in apocalyptic terms, but the Supreme Court declined to rewrite copyright law, noting that videocassette recorders had substantial non-infringing uses such as time-shifting. Earlier still, composer John Philip Sousa predicted in 1906 that player pianos and gramophones would destroy music composition, and portrait painters feared the camera would replace the brush. None of those collapses materialized. EFF points out that photography instead enabled new forms like photojournalism and a resurgence of portraiture, evidence that new tools tend to expand who can create rather than end creation itself.

Scale, and the artists already working

EFF makes two further empirical arguments. First, it cites research indicating that large generative models are unlikely to produce infringing outputs, because the more data a model is trained on, the less any single training example influences a given result. Second, it disputes the notion that AI necessarily displaces human creativity, listing artists who work with the technology: Boston-based Nettrice Gaskins, whose AI-assisted portrait of Octavia Butler is displayed at San Francisco's airport; Indian artists Prateek Arora and Varun Gupta reimagining Western science fiction; Philadelphia's Alex Smith working in Afrofuturism; MIT architecture professor Ana Miljački, who used generative AI for a documentary on Yugoslav World War II memorials; a research project amplifying voices from Iran's Woman Life Freedom movement; and the company Bronze, which builds generative music with artists including Disclosure and Jai Paul.

It is not a court's place, EFF writes, to decide who counts as an artist or which tools foster human creativity. Like the videocassette recorder, large language models are general-purpose tools, and while their effects will be far-reaching, disruptive and potentially harmful to some, the group contends that distorting copyright law is the wrong instrument for addressing those harms.

Why it matters

The lawsuits EFF is engaging with will help determine whether training and deploying generative models counts as fair use — a question with consequences for every AI developer and for every creator whose work sits on the open web. If courts adopt the market dilution theory, copyright could shift from a limited incentive to create into a lever that incumbents can pull against any competing technology, AI-powered or not. EFF's argument is for the opposite outcome: keep copyright's existing boundaries intact, and let new markets develop under the same rules that governed every disruptive technology before them.